Impact of the spatial resolution Variability within pixels: Normalized Standard Deviation (NSRDV) Number of patches per pixel Number of classes per pixel Purity (patches/classes)
Conclusion (obvious) High spatial resolution is mandatory for agriculture applications (field/species level) – Large differences between 1km and 300m resolution – But large difference also between 300m and 100m! Impact on: – classification – disaggregation – LAI estimates (non linearity)
Development of LAI, FAPAR, FCOVER near real time products from PROVA-V Objective: Develop biophysical products for PROBA-V LAI, FAPAR, FCOVER 10 days frequency Near real time Consistent with VEGETATION products (GEOV1/2) 300m resolution Impact of higher spatial resolution No climatology available Climatology/pixel meaningless No use of SWIR
GEOV3 GEOV1 MODIS Compositing: Filling small gaps
Conclusion Associated quality indicators – Flag – RMSE – Confidence interval from polynomial fit Consistency between variables – Same selection of observations used (mostly based on LAI) Importance of the compositing step – Without climatology – For short term projection Needs to be done: – Implementation of the algorithm – Verification/Validation – Checking consistency with GEOV2 (aggregation at 1km) Updating the algorithm when enough actual PROBA-V data will be available Building a longer time series: process back the MERIS archive? Combination with S2 -> 20m observations every day?